The Trump administration approved a framework that allows NVIDIA to resume exports of its H200 data center accelerators to China under a controlled licensing regime. Rather than fully lifting export restrictions, the arrangement requires recipients to be vetted through Commerce Department licensing and imposes a revenue-sharing condition that allocates 25% of sales proceeds to the U.S. government. The licensing requirement means shipments will be controlled and limited to approved commercial customers, a mechanism designed to reopen a major market for high-end accelerators while maintaining a regulatory channel to monitor distribution.
The H200 chip remains positioned as a capable solution for large-scale Artificial Intelligence workloads. It is manufactured using TSMC’s N4 process and combines NVIDIA’s ‘Hopper’-derived architecture with 141 GB of HBM3E memory and 4.8 TB/s memory bandwidth, features the article identifies as making it well suited for large model training and dense inference tasks. The H200 has been shipping since spring 2024 and benefits from a mature driver and software ecosystem, which the piece highlights as an advantage for commercial deployments even as newer architectures arrive.
The report notes that the H200 does not match the performance of NVIDIA’s current ‘Blackwell’ family and other upcoming accelerators, but describes it as a mature, competitive product whose raw compute power is roughly double that of solutions like the Huawei Ascend 910C. By combining a reopened commercial channel with revenue sharing and licensing oversight, the decision both restores NVIDIA access to a significant customer base in China and preserves a regulatory mechanism to limit distribution of high-end accelerator technology.
